Dynamical Systems Theory for Sport and Movement (Audio + Transcript) (Patreon)

• If you’re new to LMA, start with our table of contents

[This is a follow-up to "Instruction Doesn't Foster Development, It Hinders It"]

This is an introduction to dynamical systems theory and ecological psychology for sport and physical movement with Professor Andrew D. Wilson. However, since how we do one thing is how we do all things, challenge yourself to apply these frameworks to your working understanding and analysis of the world.

The topic is dense, so it may require more than one listen (there's also a summarized transcript with commentary for those in LMA). I especially encourage practitioners in the in-person LMA program to give this a listen. Since our pedagogy is so radically different from what people are used to, understanding some of its principle underpinnings will significantly improve your own LMA learning process and reduce friction.

For Marxists, these frameworks might be the missing elements in their theory. These frameworks, without Marxism, are also incomplete because they do not fully describe the world but instead isolate and decontextualize bits of the world.

Regardless of your physical movement practice, this information will both inform and strengthen your approach.

Andrew D. Wilson researches perception, action, embodied cognition, and learning at Leeds Beckett University in the UK. In his own words, "This non-representational, non-computational, radical embodied cognitive science is at odds with the dominant cognitive neuroscience approach but provides an over-arching theoretical framework that I believe psychology is otherwise missing."

The first interviewer is Rob Gray, an associate professor of human systems engineering at Arizona State University.

The second interviewer is Stuart Armstrong, the head of coaching for Sport England.

The interviews have been edited and modified for clarity.

Transcript

[Note: The transcript is heavily modified for clarity and includes my own commentary.]

Gray: What does the ecological approach or ecological psychology mean to you? What's the key difference?

Wilson: It's quite a big mindset shift from the more standard cognitive representation of information processing approaches. For me, it boils down to the place where you start. The place where you begin is in perception, and you start taking perception seriously. And that was what [James] Gibson actually did. He just realized that there were unanswered questions about how perception worked that would have consequences all the way up through the system. And so he just got really interested in trying to figure that out.

For me, the logical approach is always: what is the perceptual information available in a task that can support the behavior you're trying to explain? And if you can come up with a story where you can identify that information, manipulate it, and show that the behavior changes with perception, you're off to the races.

So the hypothesis is that most of our behavior is not preplanned or abstract or anything like that, but it's embodied and embedded in the environment. And the shapes of behaviors connect to the way we perceptually interact with our environments. That means you need to know two things. You need to know what it is about the environment you're perceiving because not everything's relevant. And then you need to know, well, how do you go about perceiving those properties?

And then you get affordances as one of the things that you're mostly perceiving. Then, you get perceptual information, specifying information variables as the way to access those affordances.

So, in the ecological approach, the trick is to take very seriously that behavior is being shaped in real time by perceptual contact with the environment.

[Sam: The conditions shape our behavior but not always directly. Sometimes, our perception of the conditions informs our behavior. For example, the condition of poverty automatically shapes behavior. How we perceive poverty, however, determines our actions to tackle poverty.]

Gray: That's a good way of thinking about it. A phrase I often see you use is task dynamics create information. What do you mean by that?

Wilson: So if you're going to tell a story in which the environment and your perception of the environment shape behavior, then the first thing you need is a theory of what is in the environment that is action-relevant. What is it that we're actually interacting with? Because not everything matters, right?

People do crazy physics to try and figure out how things are built, but most of that's irrelevant to an organism because we're too big to worry about electrons and all that sort of stuff. So the question is, what's a good way of describing the animal-sized environment? And is there a way of doing that in which things like affordances make any kind of sense?

If we're going to characterize the environment in a useful way, we have to do it at the level of dynamics. Those are the tools we're going to use to have a complete description of the ecological environment. And so what that means is, if you want to describe a task, you use dynamics.

Dynamics is just a mathematical framework—dynamical systems theory. But dynamics is a mathematical framework that contains some things and not others. So, let's talk about position and changes in position. So velocity and acceleration. So it lets you talk about change over time, but it also gives you variables like mass, which lets you talk about the forces that cause those changes, and that's what you got access to in a dynamical description of something. And it turns out that if you want to adequately describe the physical environment of an animal-sized organism, those are the tools you need, right? You have to start there. So, characterizing the environment is about doing it in dynamical terms.

And then doing it ecologically means not just describing it any old way but trying to use those tools to build a description of the environment as it's presented to an organism. And that's where you get into affordances, the idea that you don't just perceive that something is a certain number of centimeters away, but that it's within reaching distance, for example.

[Sam: Dynamical systems describe the world at the level of math. The ecological approach describes it at the level of the organism.]

The basic idea of task dynamics is that you describe one task at a time at a dynamical level of analysis, and that gives you the right description of what it is that the organism is trying to do, and that's your starting point for explaining their behavior in response to that task.

Gray: You've mentioned affordance a couple times. What does that term mean to you?

Wilson: People talk about it as an opportunity for action and all that kind of stuff, but that's quite vague. You could have all kinds of opportunities for behavior. And then the question is, well, how do you perceive them? That's actually the important thing. You have to create information. Otherwise, who cares, right?

So, in the theory of ecological psychology, there is the notion of affordance. One is affordance as a relation. So, it's a thing defined simultaneously by the organism and the thing out in the environment. And that's one way of thinking about it. And that tries to capture the idea that you can have a chair that affords sitting for me, and I can create the affordance, but a child, for example, would come along, and it might be too big for them. And so the affordance is absent for them. That's trying to capture that idea.

But that falls apart a little bit. It's being created by the interaction? How does it create information? What is it that light is bouncing off, for example? And also, how do you learn it? Right? If it only exists once you know how to do it, then how do you make it? How do you get in on that?

For me, the best way to go about affordances is to think of them as properties of the environment—the dispositions of the environment. And so a disposition is something like the disposition of salt to dissolve in water. So if you've got some salt and it's just there, it's not dissolving. But if you stick it into water, it'll dissolve because it has that disposition. It is disposed to do so. Same with an affordance.

So it's a property out there, and you find it with no reference to the organism, and it's just there, and it's interacting with light, and it's doing its thing, and it's creating information. Then, if an organism comes along that happens to have some matching properties, some complementary properties—for example, I come across my chair, and my knee comes out to be about the right size, and I can bend in the right places, then I can now interact with that affordance, and I can affect it. Now, I can enter into a little perception-action loop, which means that the affordance can shape my behavior.

So, an affordance is a property that's out in the environment and describes the physical environment in ecological terms. It gets back to the idea that you have to have a good theory of what the environment is as far as an organism goes. And affordance is a thing that's an ecologically relevant, perception-action-relevant property of the world.

There are lots of other properties of the world, and some of them just don't matter. For me, that's what an affordance is. So it's a thing you can identify.

[Sam: I can triangle choke Bob; that's an affordance I have. That affordance is absent for my 4-year-old son. Nothing about Bob has changed, but we have different affordances because my son and I are different. Affordances are not universal. My son can run underneath Bob's legs without bending over. That affordance is absent for me.]

Gray: So I thought I'd throw out a term, and you could briefly explain what comes to mind. So first: functional equivalent.

Wilson: So the idea is that there are many different ways of doing most behaviors. I guess that would be the way to go at that. For example, if you're reaching out for your coffee cup, you can do it in many different ways. So most people just reach out in a straight line, and you can just do it. But if something's in your way, you can reach around it. If your arm is tired, you can go more slowly. There's a bunch of ways that can pull it off. And the idea is that they're all equivalent because they pull off the same function, namely moving your hand to grasp the object. And this gets into the whole, degrees of freedom versus the, what's Latash's phrase? The bliss of motor abundance, which I love.

Having these options—the ability to respond to the environment in many different ways is a source of flexibility. It's not a problem. It comes with some issues, but it's not fundamentally a problem. It's a huge perk, a massive benefit of the system, enabling you to cope with what's happening in the environment.

[Sam: Whether I touch you to the face in a straight line, underneath, in a downward arc, or from the left or right, they all do the job of touching your face.]

Gray: Yeah, and that was my next one: motor abundance. I noticed you have a great video of—is it a lynx or something at a zoo trying to get down from a rock?

Wilson: Yeah, it's like a leopard—a snow leopard.

Gray: Yeah, it takes the most unpredictable action to get down off this thing.

Wilson: I know, I couldn't stop watching it. But then I was like, let's start on the assumption that that animal is not unwell and try to figure out why it might have done that. I have no idea if my analysis is true, but the idea was just, what would make you do that? And it basically boiled down to, it could do it, and right there in the moment, it was the best solution.

Gray: It made sense to it for some reason.

Wilson: Yeah. But to understand that, I had to think about the environment it was in, what were the affordances, what were the opportunities currently facing that animal that it could do, that it couldn't do? What was it trying to achieve? So I had to make some guesses, but the idea was all about figuring out the ecological shape of the environment that it's in, which made that maneuver a legitimate option. And it clearly was a legitimate option because it pulled it off without blinking.

[Sam: Motor abundance allowed the cat to reliably perform a difficult task while being flexible in the movement particulars.]

Gray: Next one: invariant.

Wilson: Oh, yeah, this one gets messed up a lot.

So when I was a Ph.D. student, one of the lessons that Jeff [his professor] really banged into my head was that there's the world, and there is information about the world. And those two things are not the same thing. You have to keep them separate in your analysis. One of the things I see about the word invariant is it gets thrown around as if it were describing something out in the world. So people talk about invariant configuration between, I don't know, rugby players as they try and do a particular move or an invariant configuration in some sport or whatever.

You can use the word in that sense if you want to, but you're not using it in an ecological sense. Ecologically, an invariant is an information. Take the optic array, right? The light bounces off my coffee cup and into my eyes. The specific little bits of light coming off there are changing intensities as I move and as I get closer to it, etc. So there's a lot of wild variation going on in terms of just the local behavior of the light. But then there are higher-order relations. There are things that remain true across those transformations. So, as I get closer, some of the details change, but the relations between angles on the object, etc., even though the absolute values change, the relationship between them stays the same.

Gibson's argument is that first of all, if you're going to look for information, it has to be something that persists, that is sufficiently stable, that you could potentially learn to use because learning takes time. Then he identified that you could have these invariant features of, for example, the optic array. And then, if you look at those, it turns out that those slightly more complicated things can be informative about interesting properties of the environment.

So if you put all that together, what you've got is that ecologically, an invariant is a relational feature of the information, and it's a thing that remains constant over a transformation, and that transformation could be you moving or could be the object moving or changing light intensity—that kind of thing.

I have this weird example from early on in the blog about right-angle triangles. So you know how you can have a right-angle triangle of any size. You can make it bigger or smaller, but in all right-angle triangles, there's a relation that remains invariant over that size transformation, and that relation is the Pythagoras theorem, right?

And so there's an invariant there, and it's only invariant over these transformations because it's a relation. The specific numbers, the lengths, and all that sort of stuff, that's all changing and changes as wildly as you want. But if it's going to stay a right-angle triangle, that has to remain invariant. And, in theory, if you built an artificial system that could directly detect the Pythagoras theorem, then it would be a right-angle triangle perceiver. And the idea is that we actually do that, just with more behaviorally relevant variables.

[Sam: You improve by training with different partners because everyone has different body shapes. Styles can change, but what remains invariant is the physics of two different bodies. By adapting to their shape, you're adapting to things that cannot change, simplifying an already dynamic task.]

Gray: Next up: degeneracy.

Wilson: So this idea is cropping up all over the place, and it's looking like it's being the rule rather than the exception. And again, it gets back to the idea of having more than one way of achieving a given function.

If you've got more than one way to do something, it means two things. First, if you try to do it one way and it doesn't work, you've got something you can pull back on. So that gives you flexibility. Second, it also points to the idea that there might not be any such thing as the single right way to do something, which I know is a topic that comes up a lot with a lot of the other guests and a lot of the sports literature about trying to train people. Then what do you train them on? It might be the case that there's no right way to do it.

But what you want is to have experience with some of that variability in order to be able to switch around as you need to. And yeah, it shows up all over the place. It shows up in nervous systems. So, my favorite example is that in the gut of the lobster, there's a killer rhythm that has to happen in the stomach to move the food along. And it's really important that this works. And it turns out that given the way the gut is wired up in the lobster, there are about 450,000 different ways of wiring up a lobster gut and still having it work properly. And that's degeneracy. And that shows how important it is not to lose access to that motion, right? So nervous systems do it, action systems do it.

[Sam: Degeneracy means I can touch someone to the chin using my shoulders and triceps, biceps and wrist, or twisting at the trunk.]

Gray: Last one: specification.

Wilson: Remember I mentioned a minute ago that you've got the world, and you've got information, and those two things are not the same thing. So we talked about describing the world in terms of its dynamical characteristics, and that's the level of analysis. So it turns out if you look at information variables—say if you look at the structure that comes off that dynamical world and gets projected into the light, it gets projected into what we call kinematic patterns, and that means you get variables like position, velocity, so any change of position, etc., and time, but no mass.

So, the optic array, for example, doesn't need mass units to describe any of its activity, whereas you need mass units to describe the forces creating that change out in the world. So, you want to be able to perceive the dynamics of the world. All you've got access to is this kinematically changing pattern in the optic array, for example. How does that work? They're not the same thing. So, how do you use one to get to the other?

And so the big argument in the ecological approach, the whole idea is, okay, they're not the same thing, but they stand in what's referred to as a one-time mapping. And that's specification. The idea is that a given dynamical event out in the world only produces certain information variables. And those certain information variables only show up when that dynamical property is present. The idea is that they trade out one-to-one, which means you can use the information to stand in for the world and have access to the world. So, it's a very important concept in ecological psychology.

It's how you get to use ecological information variables as if they were the world and have it work. Without specification, you'd go, well, I'm picking this thing up. And it mostly refers to this property being present, but sometimes it doesn't. And then you're in the world of having to make and do inferences. This is why cognitive psychology kicked off and the thing that it does, because it thinks there's no such thing as specification. It doesn't think you can get it.

And underneath it, there's a slightly complicated, slightly philosophical, detailed argument in the ecological approach that walks through how you get it. And it's just because the information is created by the laws of physics, right? And you rely on that, which is very helpful and important.

Specification is the one-to-one mapping between information and the world. And because that exists, that means that using the information to stand in for the world works. And that's why we do so well when interacting with the world.

[Sam: Even though we don't ever perceive the direct world, the information we gather is reliable enough to interact directly with the world. You get better at sparring games by playing sparring games. What you're improving is your ability to perceive and make quick reads. Rather than relying on recall to remember what comes next, it's about getting better at seeing what comes next. If we aren't relying on perception, we're not gathering specific information about our world, effectively sparring blind. This is why drills and pad work are less helpful because you are collecting information about a specific environment and then trying to remember it and shoehorn it into another environment rather than gathering information about the conditions you're currently in. You're not directly engaging with your material conditions but with your distorted memory and your imagination.

Armstrong: You've discussed dynamical systems theory and ecological psychology, and these terms are thrown around quite a bit, certainly in the coaching space, as more and more people are becoming aware of them. What are the distinctions? How do we understand the two of them? How do they interrelate?

Wilson: Dynamical systems theory is a fairly abstract mathematical formalism. So dynamical system theory is not about people in particular. It's a way of thinking about complex systems that change over time and interact with each other, and it's a way of trying to describe what's going on in those complex interactions in a way that you can get some traction on. Math is a way of formalizing that description so that you can simulate it, play around with it, and grasp it.

For me, dynamical systems theory is the right language for describing these systems, how they work together, etc. However, the ecological approach asks about the nature of the world at an ecological scale. What is out there to be perceived? What is there to interact with? What do we have to be able to do in order to be functional?

It's also a theory about how we perceive it. So you get affordances, that's what's out there to be perceived, and you get information. So Gibson provides a theory of people, or a theory of organisms actually, and how they're connected to their environment and to each other. Dynamical systems theory is the right way to formalize those ideas. But if you want to study people, you need reasons to ask certain questions and dynamical systems won't get you there. Gibson will.

Armstrong: The assumption is that, within a dynamical system, it tends to be that humans behave in a certain way due to how those systems interact. The assumption is that people behave because of these systems, but actually, what you're saying is the behavior is derived from what people perceive while they're within the system, and how the system changes their perception changes their behavior, not the other way around.

[Sam: Trainers focus on the action, the behavior, when what comes first is perception. To improve sparring, don't focus on isolated mechanics (techniques). Improve perception and design what the practitioner perceives.]

Wilson: Absolutely, that's exactly it. Dynamical systems are the right way to describe what's going on, and the ecological approach is the right way to explain what's going on. The ecological approach uses dynamical systems theories to formally describe its explanations. But the explanation, the reason why something actually happened lives in the ecological part.

Armstrong: So dynamical systems are what's happening, and ecological psychology is the why or the how?

Wilson: Fundamentally, yeah, that's exactly it.

Armstrong: Something else you just said there, which is really important, is a distinction between affordances and information. That's worth pulling out.

Wilson: There is a very important distinction between the world and the information about that world that we use to perceive it. So, coordinated rhythmic movement, where you get two fingers and wiggle them in some coordination. Same thing at the same time or alternating, in-phase, anti-phase, that kind of thing. When you're asking people to move in a certain way, there are only so many things you can do without training. People can do in-phase and anti-phase, but if you ask them to do something in between, the 90-degree one, a syncopated rhythm, they can't do it without training. So the question is why is behavior organized that way? Out in the world? There's no particular reason why those should be different, right? 90 degrees, 180 degrees. They're very simple movements and coordination of the system.

So what turns out is that out in the world, things are, you describe things dynamically, things are out there. Now the question is, how did you perceive them? And so you don't perceive relative phase when you perceive these coordinated rhythmic movements. Relative phase is the thing that's out in the world. That's the description of the behavior of the world. We don't get direct access to the relative phase. What we've got is these things produce motions, and they produce structure and light and its patterns of motion.

And these patterns of motion that we use to detect relative phases explain why the behavior is so weird. We learn to perceive things in terms of relative direction of motion. That explains why the behavior has this particular kind of structure.

The formal way to think about this is that the world is best described at the level of dynamical systems. You want to know how things are changing, their composition, what they're made of, and the forces involved in why things are changing. But those things out in the world produce patterns of change in motion in the optic array, the acoustic array, and all the various energy media we're sensitive to. Those patterns are what's referred to as kinematic. They're just patterns of motion over time, no force information. So, the world is one thing.

And it produces information about itself. Those two things are not quite identical. And our interaction is very important.

[Sam: I bring this up often: similar does not mean the same.]

We interact with information, right? That's our interaction with the world. So that information can specify what's going on in the world, but it's not the same. And so the structure of our behavior comes from the nature of the informational content that we have with the environment.

[Sam: Our actions are not directly informed by the world but by our perception of the world. So how well we interact with the world or make decisions is directly related to how well we perceive. We wrongly assume we are perfect perceivers when we take perception for granted.]

If I want to explain grasping behavior or my coffee mug, I need to be able to characterize what it's like as a physical object, but I must also characterize the layer of information it creates because that will not be identical. When I perceive this coffee mug, it's not like a little copy of the coffee mug is coming to me. There's information about it, and it's informative and specific, but it's not identical. And it's not a little copy, right? There's not a little copy of the coffee mug that ends up in my head. What I'm doing is I'm interacting with that information.

There is the world you have to characterize one way, and the information about it gets characterized slightly differently. And those two things are very tightly, very intimately, lawfully related, but they're not the same thing. And so if you want to explain behavior, you must have all the pieces.

Armstrong: About the coffee mug. Once I grasp it, I can understand something about it in terms of its form, shape, what it might be able to do, and all those sorts of things.

Wilson: And those are actually the primary things that we interact with. The fact that it's a coffee cup is just a name we've given it.

Armstrong: And so I might decide to carry it as a piece of jewelry or something like that. And again, I'll find it's different from whatever, but over time, I'll discover that there are various ways I can interact with this coffee mug. One of the things I can do with that, what it's most useful for, is putting liquid in that I can then drink from, which helps me with my survival. Are the ways of interacting with the coffee mug affordances?

Wilson: Everything you described. Affordances live in the world. Everything you just described was the properties of the object and how those properties enable you to do some things rather than other things. Those things live in the world, end of the equation.

My coffee mug is a surface that's arranged in a certain way such that if I put liquid into it, it'll hold it. It's arranged in such a way that it's a certain size so that when somebody like me with a hand about my size comes along, I could interact with it in certain ways but not other ways, and so on and so forth. All of those things that you just described are the basic properties.

[Sam: Affordances come with the coffee mug. Affordances come with each training partner. They provide the affordance. Knowing techniques in isolation does not matter all that much. Knowing how to jump on a box will not help me all that much to jump for a layup in a live game.]

The fact that it's a coffee mug is irrelevant. The fact that it's an object of a certain size, certain weight, certain material properties, those are the kinds of things the system cares about. That's all out in the world. Those properties are when they come together in the shape of a coffee mug, create the affordance of graspability, create the affordance of, you know, containability, right? This thing can contain liquid, etc.

[Sam: The styles don't so much matter as much as the body shape of your opponent. The names of your techniques don't so much matter. What matters is the properties of your movement and whether or not it complements the affordance before you.]

Affordances live in the world. Affordances are about the way the physical world is. Now the question is, how did you perceive all that? So there's a bunch of different ways. You were talking about picking it up and exploring. So, obviously, that's a haptic exploration of the object. That's still a perceptual system. Just because you're in mechanical contact with it doesn't actually give you any magical access. That's still a perceptual system.

[Sam: It's still the perception of the world rather than the direct access to the world.]

I can look at it visually. I can interact visually—guided reaching and all that, having the haptic and visual systems work together. The question is, what is my visual system interacting with? Well, my visual system is interacting with patterns in light that are coming off my coffee mug. And those patterns change as I move around the coffee mug; they change in certain ways. There are regularities in those patterns that tell me, oh, as I move around, you know, I'm getting information about the size of that object, right? Because there are certain patterns and relations in the information that don't change. And so I can now use those patterns to guide and structure how big my hand is when I reach out for it, etc. And then I can fine-tune that when I actually come in contact and so on and so forth.

So all the things that make the object afford anything, all those properties live out in the world. My job is to perceive those using patterns and light patterns and the deformation of my haptic system. So, the idea is that there's stuff in the world and I want to be able to perceive it.

As a scientist, if I want to explain what's going on and analyze the task into its actual components, I must decompose it into world bits, information bits, and organism bits. And that's where the language of affordances comes in—the world bits. [Michael] Turvey calls the organism bits effectivities. Those are the complementary bits that enable me to affect the affordance, right? So the fact that my hand is a certain size, et cetera, that sort of stuff. But the critical thing in the middle is that there's information in light, in various other acoustic, various other arrays. And that's how I perceive, that's how I perceive the affordance. And it's that information that lets me use my effectivities, change the shape of my hands, change my location and space, etc., to interact with those.

[Sam: Without this understanding, practitioners will think they can effectively mimic any style regardless of conditions (effectivities, affordances, and invariants).]

Armstrong: And so if we manipulate the organism bits, the effectivities, we manipulate the available sources of information, for example, constraining and change.

Wilson: Organisms work in real time. They interact with what they have available to them now.

And so if you alter that access, then their behavior changes and adapts, and they produce sometimes very different patterns of behavior. My favorite example of this right now is a paper I read recently by Ian Renshaw. He was looking at cricket batter behavior when they were facing a bowler versus a bowling machine.

[Sam: A bowler is basically a pitcher.]

The inspired bit of this paper was that they didn't just look at things like accuracy or how well they hit the ball, etc. What they actually looked at was the coordination patterns of the batter's whole body. So they measured a bunch of joint angles and looked at the timing of the forward step and a bunch of other things. And they found that the cricket batter, the expert batter facing a bowler or a machine, was perfectly capable of hitting the ball. But because the bowling machine removes all the information that comes off the bowler about what they're doing and what they're about to do, what happened was that in order to do it, they produced qualitatively different coordinated patterns of their body.

[Sam: How a batter swings at a ball is based on the information provided by the pitcher. Same as when you're live sparring with a partner. When you replace the pitcher with a machine or the live partner with a pad holder, all that relevant information is absent. You need to practice reading relevant information; a pad holder or a machine does not provide that. You'll train different solutions for different tasks.]

So, they achieved the same goal but did it in qualitatively distinct ways. It shows that the information coming off the bowler as they approached for their delivery is something that the system has access to normally, is informative, and is actually worth it. It's something the system has learned to use to shape its behavior. In the absence of that behavior, it's fine. It can still hit the ball, but it has to come up with a different solution. It has to time things differently because it can't time itself now relative to the behavior of the ball.

So it's a great example of, at one level, they did a lot of really clever things to make sure that the behavior of the ball coming out of the machine was very similar to that of the ball coming out of the bowler's hands, for example. They did a lot of very cool stuff. And it turned out that that wasn't what the system was interested in. The system was interested in a much more expanded thing, and the moral of the story was exactly as you were saying: the bowling machine is fine at one level, but it does create an aperture. It does restrict access to some information about some affordances, and the system was like, okay, that's fine. I can still behave and produce an outcome, but I will have to do it in a very different way.

[Sam: The system wants the whole story, not an out-of-context sentence from the story.]

So now you get into questions of transfer. If you spend all your time training, basing against a machine, and then you go out on game day and base against a bowler, you haven't had any opportunity to learn any of that information about the bowler. And so, with all of your behavior, you're still going to be able to try and hit the ball, but you're going to be doing it based only on information about the flight of the ball. Of course, in cricket and baseball, you've only got 400 milliseconds, and so on. So there are consequences here.

Armstrong: So, as coaches, we are constantly trying to use this information that we're getting from the kind of theoretical space as a means by which to help us identify better, more effective approaches to help our athletes perform. So this transferability question becomes really essential. Some coaches might say that a bowling machine has a function for us. But we must understand that function won't necessarily be immediate transferability.

Wilson: The trick is to remember that time is a big chunk of this, right? Things happen in different orders and produce different behaviors, right? Time and ordering are significant. I don't know if this is the right way to go, and I don't know what the best order would be. But one way might be that you bring your batter in; he's not so good right now, they're still developing some technique issues, and there are still some basics that you want to work on. It might be the case that spending a little bit of time with a bowling machine just means that you go, look, you know what, you're going to have to learn about what bowlers look like at some point because that's actually quite useful information. But right now, I want you to focus on something else. We're going to strip that out for the time being. We'll spend a little bit of time like this, but then I think you're exactly right. The trick is remembering that the bowling machine is different from the game context. It doesn't overlap in some interesting ways.

So, you might not get an immediate transfer, but using the bowling machine could be a tool just to get the batter to a certain point where the basics of their technique are there. And then you go, okay, now we can skip that. Now that you've got, you know, you know how to swing a bat. Now, we can start exposing you to the behavior of bowlers so that you can start learning the information that comes from them and how to use that in the control of the swinging bat. And then you might expect more transfer.

So it's an old problem. So the question is, if you've got your game context, your full task-rich environment, all these kinds of things, the age-old question is how do you break it down into components so that you can train those components without violating the basic rules or the essential relationships and the fallback?

Yeah, you can't. I think people tend to agree. You can't just throw someone into a full game. It's too busy. There's too much going on. You have to walk people into it. The question is, how do you do it in an organized fashion? There are not a lot of good answers right now. I'm starting to explore this now using coordinated rhythmic movement. We're using it to develop some intuitions about what the hell is a task? How can you break it down? What constitutes a natural part of that task? And how could you do that? So we're using the ecological approach now to identify those parts. But it's a very complicated question.

Armstrong: If we're saying that kind of information is arising from what we're perceiving, the things out there in the world, what we're saying is that what we want to try and do as much as possible is to try and allow the learning to take place from something as close as possible to what we're likely to perceive. I would rather see if the movement solution can emerge from the perception we're receiving rather than saying there's a movement solution that will likely solve the problem in the future. And that's, I think, a fundamental challenge that many people struggle to grasp.

[Sam: Giving someone vitamin C pills will not prepare them for the orange. Likewise, reducing a task to a mechanical reaction will not prepare someone for the totality of the task. A task is not just a mechanical movement. The movement is the reaction. Most of the task is correctly perceiving the task, like perceiving the orange or coffee mug or a baseball batter perceiving a fastball vs. a curveball. The perception drives the action. The action does not exist in isolation. Practicing the action in isolation does not even form the relevant mechanics for the task. Learning to swing against a machine does not create the same swing you would get against a person. They're both swings but not the same. Since the machine or a tee ball stand necessitates swinging and hitting a ball, people assume this is the same as a live situation when it is developing a parallel skill that is highly customized for its specific task. The skill you develop in a live situation is reading your opponent's action rather than just swinging a bat independent of anything else. Just as mechanically smacking your lips alone will not prepare you for kissing someone else, drilling movements is not preparation for game speed.]

Wilson: But that's the key to the ecological reframing of the problem right there in a nutshell, right? It is the recognition that behavior emerges in real time from the flow of information you're interacting with and that flow of information is generated by whatever is happening in the world. And if what's going on in the world is critically different in training, it's critically different from what's going on in the world in a game context, then the behaviors that emerge in training will not necessarily be relevant to the behavior. It will not be a behavior that will necessarily emerge in the context of a game situation. So, thinking about Ian's thing with the cricketer, it's entirely possible that you could use the machine to produce some technique improvement, but it's not like there's a correct technique, right?

And Ian's data shows that what you learn interacting with a ball machine is one coordination pattern over your whole body. And what you perform in the presence of a bowler is a different coordination pattern. They still, at the end of the day, produce a cricket shot, but the pattern of performance that is produced is very different. And if you want to learn, engage with, in real-time, what's currently present.

[Sam: What do you want to get good at? Drilling or sparring? They are two different specific tasks, and training in one is actually at the cost of the other.]

And so if you train people in a context that is missing some critical element, then they will learn to solve a problem, but they'll come up with a solution that depends on things that were present in training. Transfer doesn't happen very often. Transfer of learning does not transfer very easily. Learning is spectacularly task-specific, which is a little weird, I think. Presumably, that's strange to say to a sports coach because the whole point of coaching is to generate a transfer. What we're finding though, is that what you learn is what you do. And what you do is learn how to interact with a very specific task dynamic, right?

The reason why it's so specific is that the dynamical properties of the training environment produce some information variables and not other information variables. And what you learn to do is learn to perceive task dynamical context using the information it presents to you. Suppose those information variables are not present in the transfer task—in the task you're trying to transfer into. In that case, you don't have access to that information, which means that if that information is not available, then whatever you use that information for, the behavior to produce in training, you can't do it because you're not learning how to bat, you're learning how to engage with the environment, right? And so what we're finding is that transfer is mind-meltingly task-specific.

[Sam: Preparing for the task means preparing for everything present in the task.]

Our hypothesis that we're developing is that the thing that governs the transfer of learning is whether or not the tasks overlap in the information that they present to people. And the way you present information in a real-life context is that you must have the right dynamical elements. You have to overlap; tasks must overlap in the dynamical elements they contain, right?

So, you can't train soccer players to pass a ball successfully through a closing aperture of opposing defenders. You can't train them to pass a ball and expect that to transfer because what they're learning is they're just learning how to pass a ball between themselves, but passing a ball, and passing a ball through a closing gap defined by two defenders are two entirely different task dynamics. There's information about the aperture that you need to learn how to perceive and use to time your behavior, etc.

Somebody trained people to balance on a slack line and a rigid beam. We're teaching them to improve their balance. Capital B Balance. You have to interact with that task dynamic in very different ways to produce functional behavior. So, learning on one, you're not learning to balance; you're learning to interact with a very particular task dynamic, a slack line, and you produce balance on a slack line. So it's quite a deep point, but it's the idea that if things overlap in their dynamical properties, they will also, just by the laws of nature, they'll overlap in the information that they present, and then you will expect some kind of transfer. So the goal now is to try and design something, training content that overlaps meaningfully.

Armstrong: Learning does not transfer very easily. You can definitely feel that. And lots of us have fooled ourselves into thinking it does transfer. This is one of the reasons why technique-led coaching is so prevalent and also so persuasive.

[Sam: What's antithetical to being practitioner-led is to be technique-led. Unfortunately, centering learning around instruction rather than the practitioner's learning process is the societal norm, so practitioners will often ask me to put the technique before themselves—unwittingly making it top-down and rigid and becoming their own worst dictator by shoehorning techniques, whether they fit or not, rather than allowing things to come naturally.]

I think in a skill acquisition context, technique-led coaching is fool's gold. And the reason I think that is, is because I think we, as coaches, are desperate to see improvements and performance, and usually, we want to see it right before our eyes because that gives us a sense that we've achieved our tasks—giving the performer a sense that they've achieved what they came to achieve. And so we'll base it on a task-specific learning context. We say, here is a task, but it is devoid of information and perception. So it is massively simple, right?

And through either instruction or whatever it might be, you can become more effective at that task. Going back to passing the ball back and forth, can we do it more often, more repeatable? Yes, enough goes at that will enable us to do that, right? We've been successful. The problem is the assumption. We feel like, right, job done, finished. And I can get that because, for me, you'd feel like that worked for you. The problem is that what we're not doing is we're not deriving the task from that performance context. And I learned this through bitter experience, watching athletes basically disregarding things we'd worked on in our training context in a game and screaming and shouting on the side, "We've just worked on this!" You know, it's the search for answers. Why? What is it? Me? Is it? Yes, of course, it's me.

Wilson: They literally can't help themselves, right? They're in one context. But they learned in that other context. So they're now trying to figure it out. Yet, in all fairness, a big chunk of this is our fault. By our fault, I mean the research community. The motor control literature, since the 60s, right through the 70s, 80s, and all that sort of stuff, the basic motor control literature, was all about identifying the scheme, the motor program.

If we train you to swing a bat, you, therefore, now know how to swing a bat, and you should now be able to deploy that skill in a variety of contexts because you have this core competence. And so that idea that what you were training is a core competence on a core capacity that you could then deploy, and context was just noise, was the driving principle for motor control. It still is for a lot of the motor control literature in particular.

The ecological flip that happens is that context is not noise; it's signal. So instead, when you are training somebody, you're not training somebody to produce the core skill or the core capacity just because you could name it. Oh, look, I'm going to teach them to balance. It's like, well, just because you've got a word for it doesn't mean the perception-action system cares, right? So what it's doing is not learning a core capacity. It's learning how to skillfully engage with a particular dynamical context. And it's doing that based on learning to detect and use the relevant information variables to interact with that task dynamic.

So behavior emerges in real-time from the interaction of an organism with certain capabilities, interacting with a task dynamic, offering certain opportunities. And so what you're learning then is, it's always behavior in context. Context is signal, right? And so that's it.

So the issue then is, okay, we had this idea that you could decompose a task into some parts. I want to teach you how to balance on a slack line and on a beam. So I'm going to decompose it into balancing and stuff like that. And the moral of the transfer of literature is that we've been carving tasks up in the wrong place, as evidenced by the data, right? I train you on balance; nothing happens, no transfer at all.

So, I think coaches are looking for at least a set of principles for identifying what constitutes a meaningful piece that you could potentially train in isolation and expect some sort of transfer. Because it's not impossible, right? You don't have to train on the full context to get improvements. You just have to break the task at its joints. I always ground this in the empirical data around coordination, but just because it gives me a nice concrete place to think.

The relevant structural overlap that the perceptual system cares about to generate a transfer of learning is: do the two tasks share informational content? You get informational content if you've got the same dynamical task, dynamical elements present in the two tasks. And that's it. That's how you get transfer.

It's almost at one level, it's not transfer. It's not transfer of learning. You've trained in the right context, presented people with all the relevant information, and then put them in a situation where all that stuff exists but also has a bunch of other stuff going around it. So it's almost as if learning hasn't transferred. What you've done is you've successfully transferred the context, which is a whole different way of thinking about it.

[Sam: Rather than transferring skills to a thing, you're just doing the thing. No transfer needed.]

What we're finding is that learning doesn't seem to transfer. What you learn how to do is how to skillfully engage with the environment under certain circumstances. And then, if those circumstances work in the other task, you're golden.

[Sam: You learn what you learn, not learn for something else. This is why self-defense is misleading because you're learning to prepare for something else when you can only learn what you're directly learning. Learning to pretend to be John Wick will not actually make you John Wick. What you're developing is acting. Acting is a direct skill. It's the rehearsal for the thing. Rehearsing to be a doctor does not make you a doctor; it makes you an actor.]

Armstrong: I was talking about this with somebody the other day. A baseball hitting coach. He's based with the Baltimore Orioles. I think baseball batting coaches are facing a similar challenge in trying to change the behavior of athletes to embrace different practice methods, particularly in that professional domain. There are so many challenges.

It's like a kind of task design lottery, right? So what coaches are doing is they're going, right, we're going to give you some experiences of certain tasks, which we think, if the stars align, there'll be a moment in time which will be almost exactly the same, and you'll be ready for it, and it will all take place. And that's how it's been working. And when you frame it in that sense, you immediately just think to yourself, well, okay, there's absolutely no way we can now consider skill development without some contextual variable in place. Task design becomes absolutely critical.

And I think most of us are manipulating task design on the fly, based on the behavior I'm seeing, or at least how the interaction takes place. For some of us who've been quite ordered and structured and are used to seeing a logical progression of our activity towards a natural endpoint, it feels really uncomfortable and messy, and it's like you don't know if that worked. And so that becomes quite a challenge.

Wilson: So yeah, it's it's a tough one. What's clearly lacking right now, is a way of usefully articulating what a task is. Transfer is premised on the idea that you train on something and test it on something else. And the idea that those tasks overlap in some meaningful way. The whole idea is you get to meaningful overlap. But the question, it turns out, that nobody's ever been able to figure out the answer to is: what the hell is meaningful overlap? Overlap in what?

[Sam: Since martial arts sparring and combat sports are one-on-one, the only equipment is for safety rather than features of the game; it has far fewer variables and is much simpler than other sports for training design. We can play the full game.

You know, two people kicking a ball, passing it to each other. Well, that certainly happens in a game, but it turns out that doesn't actually help you pass through the aperture for the thing. So, you need to find a way to decompose tasks. There is no such thing as a single right answer that will apply across many contexts.

So I've been studying long-distance throwing. And so the basic idea of throwing is to move to produce a projectile motion. A ball, once you let it go, that's it. That's your last chance to interact with it. Then, the ball has to cross some space, and whatever happens is just a function of local physics. I can characterize the projectile dynamics the person has to interact with to successfully hit a target of a given side at a given location. Then, I can interpret their behavior relative to that dynamical description.

There is such a thing as throwing in the sense that every throw is an attempt to interact with the dynamical description, but a throw to a given target involves changing the parameters, right? So you have to throw a little faster or a little slower or a little higher or a little lower. The two tasks are both throwing at the level of the dynamical description. But at the level of if you describe the individual behaviors, you get two very different-looking behaviors.

[Sam: Important distinction, throwing and throwing to hit a target are not the same. Likewise, doing a technique in isolation and doing it to directly interact with a live target are two different things.]

I think now, in order to be able to define a task, you have to be able to find the right level of analysis where the two behaviors, the repetition without repetition, you have to find the level at which those two things are in some meaningful sense, the same thing. They are the same at the level of interacting with the dynamical description of the environment.

If you look at the details of the execution, my argument is that the perception-action system interacts with the underlying dynamics. And so this then connects to all the motor abundance, synergy stuff, the idea that there's no such thing as a single correct movement. Instead, what you're doing at any given time is the repetition without repetition that you are trying to interact with, and your behavior is being structured by the fact that you're interacting with the same task dynamical structure.

[Sam: Repetition without repetition is that you can repeat a movement 100 times, but all 100 times are slightly different, so you can never actually repeat the same movement even if you want to. Since you can't ever repeat the same movement twice, why make that the goal of practice? If the lesson is that there are various ways to do the same thing, why not practice that?]

But because of noise, error, and all those little bits, you must make minor corrections and do everything. Everything is slightly different every single time, but it all achieves the same goal. But the idea then is that what you're actually trying to teach in teaching throwing is flexible engagement with that task dynamic. And once you've got that, if that task dynamic from the game context is present in the training context and you teach people to flexibly engage with that task dynamic, then you'll get your transfer. And so the way to identify those pieces is at the level of dynamics.

Perception-action systems interact with the world at the level of dynamics through the information that those dynamics create. If you want to carve a task up in a behaviorally relevant way, you must do it at the level of task dynamic. If the overlap doesn't exist at that level, the information will be different.

Armstrong: If you take sports like cricket, golf, all those sorts of things, there is an obsessive drive towards aesthetic, technical models. And the assumption is that the aesthetic, technical model equates to performance or skill and performance.

[Sam: Also true in martial arts. That movements should look like they do in the movies or some way you've imagined, and if it looks like that, if it looks geometric or unordinary, it somehow means effective or proficient. It's substituting martial theater for sparring. Aesthetics for performance is actually a false perception of excellence. At this point, we know what a fast race car is. Wanting to add fins to it won't make it faster, nor does it look good to anyone but you.]

I think it's based on a belief that if you can make that organism perform in a certain way, kind of make it as machine-like as possible, that translates into performance. It's seductive. So, the training methodology is all derived around the idea of helping the performer fit into an ideal technical visual form. And skill, to this day, skill and actually selection and performance are predicated on how people look.

So I've genuinely seen in cricket, particularly, somebody who looks right, regardless of whether they get no runs, but they look right, will be given far longer time before they are deselected. But that drives this fundamental behavior around our job as coaches to get you to become as aesthetically pleasing, technically, toward orthodoxy as possible. Then, people like Lasith Malinga come along and break the mold. Then, all of a sudden, our intuitions are broken. And we're like, well, what do we do there?

So I've got quite a few kids; they're under 11. You know, fielding and all these sorts of things. Now, I don't do much in terms of what you might call decoupled fielding practice. But what I try to do is games where there might be a lot of opportunities to field. So, for example, quite a few hit-and-run type games. So the batter isn't getting a lot of decision-making there because they've got to hit and run, but the fielders do. And what's very interesting is, like, we watch matches, and I see kids, and they pick up the ball, and what they're getting is that my task is to sling this projectile as hard as I can. They're doing it almost like a discus motion because it feels intuitive to them and is the best means of developing power.

But if I were to reframe the task dynamics by framing it differently, giving it different information sources by manipulating what's available to them, by saying, for example, you have to throw this ball and hit that objective, otherwise you're out. I wonder if that would change the movement pattern they chose.

Wilson: So, in my not-very-illustrious cricket career in high school, in order to generate that kind of power, I did something similar. One of the things that's very interesting about throwing is that to get an accurate throw, you have to land your release within a very tight launch window that's defined by the dynamics of projectile motion. That window, time-wise, is tiny. It'll be on the order of a few milliseconds. If you cock it up on either side, then you'll miss, right? The timing requirements are astonishing.

So, with an overhand throw, if you miss it a little bit, all the error is in sort of the vertical dimension, which means you get basically into the right plane of motion, you get it in the right place, and then your fielder can do a bit of an adaptation to it, like catch a high one or a low one or whatever. But if you cock it up horizontally, right, then the errors are way away; they're too far. Like that tiny little angular difference corresponds to a huge distance by the time you get there.

So, why are they doing that? Well, one reason is they're trying to generate power because they want to cross the distance, and if you're throwing under time pressure, you need to throw as flat as possible to minimize the flight time. Throwing flat means you need to have a very high release velocity to cross the distance, right? The way to cross the distance is to throw it very high with a lower velocity but, obviously, much longer flight time, which means you don't get that run out, right?

So, what are they working on? They're working in a context where the time pressure is forcing that, or it means that of the two possible ways of crossing the distance, high and slow, flat and fast, flat and fast is the most behaviorally relevant. But fast is hard. The increased speed makes it harder to land in the launch window. Generating that speed requires careful, precise coordination as you take all the energy you store in the trunk and whip it down the arm. There are many opportunities down that kinetic chain to lose energy, etc. So, high-quality throwing requires this incredible coordination. That coordination is really hard. And if you're nine and 10 and not as big, you'll try and find a way.

So, you're trying to reframe the question as one about accuracy. Take the time pressure off, maybe initially, and get them back into the overhand throw, where all the action is in the vertical plane. It's better because smaller errors can be accommodated more easily at the other end.

I play softball and it took me a long time to learn to throw a ball properly. And I had to spend a lot of time just practicing it. But then, of course, the problem is in the game context; the time pressure will be back on, right? So you're going to see that even in training, they might be starting to get it, but under the pressure of the game context, the whole thing will disintegrate in the game context under the time pressure, right? So, it's the right way to train them, but it's going to take longer, and it's going to be harder. The performance will be more variable for a longer time, and that has motivational consequences, all that sort of stuff.

Armstrong: Yeah, which is one of the reasons why I'm trying to develop it within a game form. So that there's the enjoyment factor of the game form. I'm trying to find a means by which I can focus on the task by reframing the task and using appropriate constraints. Because at the moment, they just got attuned to the idea that speed is everything.

Wilson: Yeah. Well, time pressure. We're trying to get the runner out, right? So, that immediately creates time pressure. As soon as you've got time pressure, you want to get high speed. As soon as you get high speed, high speed is very difficult in a throw. First, I would try to take the time pressure off.

I like to create a game context where, in fact, you get them out not by getting the ball there on time but by getting it to a certain location. The better route. For example, if the ball arrives in a location where the fielder can catch it without moving a foot.

There are no general principles for identifying how to balance these things out, but creating a balanced, varied environment where the basic pressure is on accuracy but where improving speed is certainly a thing that you can do to achieve the task goal.

You must get them in the ballpark before you can start honing them in. So you've got to create a situation first where their preferred action solution is the one that will be more functional later, further on in time, and further on in training. So you've got to get them in the ballpark first. So there's an order. Again, time is a factor, literally.

Armstrong: I imagine the timing window on a horizontal action versus a vertical action is even smaller.

Wilson: Quite possibly, but worse, the consequences of missing that window are larger independent of everything else. There's a friend of mine who plays softball. He's got a very solid arm, a very reliable thrower, and he does have a horizontal arm swing. But when he does it, he's so good at it. He has got good enough at it now that if he's pitching or standing at shortstop, he can whiz it into first base reliably and at a very high speed, which means he crosses the distance and gets the runner out. So he has certainly found a way to make it work. And his errors are very rare. So even though the pressure is very high, a tiny error just becomes very magnified because of the consequences of errors.

Functional, more adaptive, and more flexible will enable more successful flexible interaction with a task environment, but it's not the case that the sidearm is just intrinsically terrible. It is just a motion that comes with certain constraints. It has some limitations and ups the demand for being accurate. But it doesn't up the demand to the point that's impossible.

[Sam: There is no right or wrong move. However, some moves come with more constraints, while others have more affordances. For example, a jab might have more affordances than a side kick. However, a jab can't always do what a side kick can.]

Armstrong: I feel my role is to allow these young developing human beings who are learning to have enough experience with various movement forms to begin to understand when one might be suitable over another. So for example, I can see a moment where a horizontal arm swing is exactly the right movement. For instance, in a rapid run-out context, right? A quick single, I've got to pick up the ball; I haven't got time to get into a vertical position.

So I've got to sling it on the move, probably while I'm falling over, and still deliver that with a degree of accuracy. That movement pattern is probably really suitable in certain situations. And it's worth the risk of it going wrong.

[Sam: You see this in modern football. Adaptive throws. Strikes are projectile movements leaving the body, so they too should be adaptable.]

Wilson: You want it in your repertoire because it's adapted under some circumstances.

Armstrong: So my intuition is to go, oh, no, don't do it that way. Do it this way because this is an appropriate movement form, but I need to disabuse myself of those. Because what I'm saying is that, what you're doing there is fine. But we want to find all the options available to us to propel the object with sufficient accuracy and speed, given a set of circumstances. So that's one thing.

Just coming back to the other point about schema theory, that the role of coaches is to try and create optimal movement patterns. The whole concept of fundamental movement skills is predicated on that idea. I think the role of coaches is to help athletes or sportspeople become more able to derive appropriate information about the task based on what they perceive. And I think the problem often with athletes is that they're not attuned to that. They're not seeing things, or they're not deriving information because, more often than not, they're actually doing one of two things. They're mindlessly in performance mode, or they've been given various movement solutions that they're trying to fit into a contextual box, hoping that eventually work for them.

[Sam: What I call recall. When practitioners haphazardly try to recall something they drilled and rehearsed.]

Wilson: You know, your analysis was spot on. It's not that they're willfully trying to do something inappropriate. They are trying to skillfully engage with the task you set them up for under their own effectivity constraints. And also, they've come with being able to do this one thing more easily. So, of course, when you're trying to succeed, and you're trying to throw it fast, and you're trying to look like you can do something, you go to the thing you've already got some capacity to do, right?

[Sam: This is why I avoid teaching parries until later or altogether because people already come in with the tendency to want to do everything with their hands, whether strike, block, or grab. It's all they'll go to at the expense of everything else, but it constrains the development of new movements. Constraining that behavior opens them up to learn so many other behaviors. Parrying will already be present even without being taught because wanting to use your hands to solve a situation is already conditioned in modern people. Parrying, like the sidearm throw, is also a very specific solution rather than the primary.]

Because of the cost to training, overhand throwing is bloody hard, right? It's really difficult to learn how to do this. That sidearm motion is natural because it removes some of the pressures. Places pressures in other places, but it takes some of the production pressures off.

The whole point of this is to identify. Once you identify what a task is, from the point of view of a skilled organism, what it does is it allows me to carve the task a little bit at its joints.

[Sam: The vertical throw is more powerful and faster than the sidearm but takes longer to develop. If you only work on the sidearm because it's initially the fastest, you'll never develop the faster throw. There's a time and effort cost to training. Spending time on one thing is at the expense of other things.]

You hear many stories about people who become really amazing athletes, and they become amazing because they tried something a bit wacky, right? Like the Fosbury flop. And possibly, Fosbury wasn't a particularly great high jumper. He was good but wasn't near the top of his field. Then one day he invented the Fosbury flop and suddenly because of the nature of that technique, was able to jump a whole bunch higher, and then that got him to the Olympics, and then I think he won and then four years later everybody was doing it and so all the better people had now just overtaken him, but it was just kind of interesting that initial innovation by him.

But innovation absolutely can pay off, right? And there's a ton of stories around like that. And if the one thing we know now about movement patterns is that there is simply no such thing as the correct technique. What there is, is the ability to flexibly and skillfully interact with the task dynamical environment. And that's where all the people doing uncontrolled manifold analysis and those kinds of analyses focus on how the movement system constrains variability.

Variability is very much part of the system. And some of that variability can be useful. Sometimes, it can take you away from the task, though. The whole idea is that you try and clamp down on variability in the performance that's screwing you up, and you try and preserve the variability. Or you don't care about the variability that doesn't interfere with you, and what that good variability gives you is space to maneuver, so as you're executing the movement, you've got to make a tiny little correction. There's room in the system because there's more than one way to do anything. So flexibility, multiple solutions, and adaptivity are all features. They're all deep features, not bugs. It's the nature of the system.

Armstrong: Kind of like norms. So there are a lot of conventions in all sports and I'm trying to break some of those conventions. In hockey, for example, the idea is it's more effective to carry the ball on one side of the field with a flat part of the stick on one side because the implement is designed to do that. And that's the most effective way of using the implement. But going back to coffee cups. Well, actually, what if you have it on the other side of the body? Yes, you've still got to use the flat side. But why is that a less effective way of doing it?

In some contexts, when you've got a defender over this side of you, that's a really effective way of carrying the ball. But we teach everybody to carry it on this side. What's wrong with having it on this side? Well, it's just harder. What if it's that we've just had less goes at it?

Wilson: Absolutely, and even if it's hard, even if it remains more difficult, having that trick up your sleeve suddenly makes you a more flexible adaptive player. When new tricks come up in sports, right? You see this every now and again, right? Say all your hockey players had added the knack of carrying the ball on the opposite side, and that wasn't very common across hockey. The first time that shows up in a game context, you will be completely bamboozled by the opposition because everything about their defense plan and defensive strategies is premised on the idea that the ball will be over here. What's going to happen is that you're simply going to reverse everything, and all of a sudden, nothing they do is now afforded by the environment.

But then you get into all kinds of interesting arms races. You see this in sports all the time. Someone comes up with a new offensive trick, and it works for a while, and then the defenses start to adapt, and so you get into this very interesting kind of arms race. Nobody teaches the details of that new trick or defensive response.

It's not about perfect technique; that's not how you do it. What you get is that the people who do well are the ones who can flexibly adapt. You see it all the time. You see that one trick that comes out for a little while that works really well, but then it's really interesting to watch the game adapt around it, and it adapts in a very unguided way. It's not a clearly specified solution. People try stuff out, manipulate things, figure out what's going on, and there will often be a good solution that everybody converges on.

[Sam: Common moves arise in sparring because those are the reliable solutions practitioners converge on. However, some of that is due to tradition. We don't want that bias.]

There was a good one in baseball a little while ago. I can't remember what it was, but something changed in the game. Then, the fielding configuration had to change a certain way, and everybody settled on this one solution. Still, it took a little bit of time, and it was very much a bottom-up kind of exploration process that had to happen, and then everybody had to learn how to flexibly engage with the environment in that new configuration because it changed. So that's it.

[Sam: It's called the infield shift. You see this often in Brazilian jiu-jitsu; new moves come and go. In the privatized sport of Brazilian jiu-jitsu, the Fosbury flop is the rule, not the exception.]

I think the most important thing out of the dynamical and ecological approach is just that recognition that time and the order in which things happen is a critical defining characteristic. That's the dynamical system understanding. Time matters. Things take time. Things happen. Systems respond. Some of those responses, the nature of those responses, everything's talking to each other and feeding back into each other and all that sort of stuff. But the order in which stuff happens matters.